Refactor API to use Azure Blob Storage instead of S3; update requirements and remove unused environment variables.

Dockerfile.api

@@ -27,10 +27,6 @@ COPY examples ./examples
 ENV PYTHONPATH=/app
 ENV HF_HUB_CACHE=/app/checkpoints/hf_cache
 ENV TORCH_HOME=/app/checkpoints
-
-ENV AWS_REGION=us-east-1
-ENV S3_BUCKET=elevenlabs-clone
-ENV S3_PREFIX=seedvc-outputs
 ENV API_KEY=12345
 
 EXPOSE 8000

api.py

@@ -4,7 +4,8 @@ import uuid
 from contextlib import asynccontextmanager
 from tempfile import NamedTemporaryFile
 
-import boto3
+from azure.storage.blob import BlobServiceClient, generate_blob_sas, BlobSasPermissions
+from datetime import datetime, timedelta
 import torchaudio
 from fastapi import BackgroundTasks, Depends, FastAPI, Header, HTTPException
 from fastapi.security import APIKeyHeader
@@ -39,22 +40,35 @@ async def verify_api_key(authorization: str = Header(None)):
     return token
 
 
-def get_s3_client():
-    client_kwargs = {'region_name': os.getenv("AWS_REGION", "us-east-1")}
+def get_azure_blob_client():
+    account_name = os.getenv("AZURE_STORAGE_ACCOUNT_NAME", "getpoints")
+    account_key = os.getenv("AZURE_STORAGE_KEY", "ts/PL1cr3X1F9JWgksAtqcWsQvPBK9UJ3BtNQBL98kYU17U3JxEiFI2vJrNDzmAyFRleOdRdoG03+ASt9RDnZA==")
+    blob_endpoint = os.getenv("AZURE_BLOB_ENDPOINT", "https://getpoints.blob.core.windows.net/")
+    blob_service_client = BlobServiceClient(
+        account_url=blob_endpoint,
+        credential=account_key
+    )
+    return blob_service_client
 
-    if os.getenv("AWS_ACCESS_KEY_ID") and os.getenv("AWS_SECRET_ACCESS_KEY"):
-        client_kwargs.update({
-            'aws_access_key_id': os.getenv("AWS_ACCESS_KEY_ID"),
-            'aws_secret_access_key': os.getenv("AWS_SECRET_ACCESS_KEY")
-        })
 
-    return boto3.client('s3', **client_kwargs)
+blob_client = get_azure_blob_client()
 
+AZURE_CONTAINER_NAME = os.getenv("AZURE_CONTAINER_NAME", "seedvc-outputs")
 
-s3_client = get_s3_client()
 
-S3_PREFIX = os.getenv("S3_PREFIX", "seedvc-outputs")
-S3_BUCKET = os.getenv("S3_BUCKET", "elevenlabs-clone")
+async def ensure_container_exists():
+    """Ensure the Azure container exists, create if it doesn't"""
+    try:
+        container_client = blob_client.get_container_client(AZURE_CONTAINER_NAME)
+        container_client.get_container_properties()
+        logger.info(f"Container '{AZURE_CONTAINER_NAME}' already exists")
+    except Exception:
+        try:
+            blob_client.create_container(AZURE_CONTAINER_NAME)
+            logger.info(f"Created container '{AZURE_CONTAINER_NAME}'")
+        except Exception as e:
+            logger.error(f"Failed to create container '{AZURE_CONTAINER_NAME}': {e}")
+            raise
 
 
 @asynccontextmanager
@@ -62,8 +76,8 @@ async def lifespan(app: FastAPI):
     global models
     logger.info("Loading Seed-VC model...")
     try:
+        await ensure_container_exists()
         models = load_models()
-
         logger.info("Seed-VC model loaded successfully")
     except Exception as e:
         logger.error(f"Failed to load model: {e}")
@@ -77,8 +91,8 @@ app = FastAPI(title="Seed-VC API",
               lifespan=lifespan)
 
 TARGET_VOICES = {
-    "andreas": "examples/reference/andreas1.wav",
-    "woman": "examples/reference/s1p1.wav",
+    "male": "examples/reference/s1p2.wav",
+    "female": "examples/reference/s1p1.wav",
     "trump": "examples/reference/trump_0.wav",
 }
 
@@ -88,6 +102,18 @@ class VoiceConversionRequest(BaseModel):
     target_voice: str
 
 
+def download_blob_to_temp(blob_name):
+    temp_file = NamedTemporaryFile(delete=False, suffix=".wav")
+    blob_client_instance = blob_client.get_blob_client(
+        container=AZURE_CONTAINER_NAME,
+        blob=blob_name
+    )
+    with open(temp_file.name, "wb") as f:
+        download_stream = blob_client_instance.download_blob()
+        f.write(download_stream.readall())
+    return temp_file.name
+
+
 @app.post("/convert", dependencies=[Depends(verify_api_key)])
 async def generate_speech(request: VoiceConversionRequest, background_tasks: BackgroundTasks):
     if not models:
@@ -107,39 +133,46 @@ async def generate_speech(request: VoiceConversionRequest, background_tasks: Bac
         output_filename = f"{audio_id}.wav"
         local_path = f"/tmp/{output_filename}"
 
-        logger.info("Downloading source audio")
-        source_temp = NamedTemporaryFile(delete=False, suffix=".wav")
+        logger.info("Downloading source audio from Azure Blob Storage")
         try:
-            s3_client.download_fileobj(
-                S3_BUCKET, Key=request.source_audio_key, Fileobj=source_temp)
-            source_temp.close()
+            source_temp_path = download_blob_to_temp(request.source_audio_key)
         except Exception as e:
-            os.unlink(source_temp.name)
+            logger.error(f"Failed to download source audio: {e}")
             raise HTTPException(
                 status_code=404, detail="Source audio not found")
 
         vc_wave, sr = process_voice_conversion(
-            models=models, source=source_temp.name, target_name=target_audio_path, output=None)
+            models=models, source=source_temp_path, target_name=target_audio_path, output=None)
 
-        os.unlink(source_temp.name)
+        os.unlink(source_temp_path)
 
         torchaudio.save(local_path, vc_wave, sr)
 
-        # Upload to S3
-        s3_key = f"{S3_PREFIX}/{output_filename}"
-        s3_client.upload_file(local_path, S3_BUCKET, s3_key)
-
-        presigned_url = s3_client.generate_presigned_url(
-            'get_object',
-            Params={'Bucket': S3_BUCKET, 'Key': s3_key},
-            ExpiresIn=3600
+        # Upload to Azure Blob Storage
+        blob_name = f"seedvc-outputs/{output_filename}"
+        blob_client_instance = blob_client.get_blob_client(
+            container=AZURE_CONTAINER_NAME,
+            blob=blob_name
+        )
+        with open(local_path, "rb") as data:
+            blob_client_instance.upload_blob(data, overwrite=True)
+
+        # Generate SAS URL for temporary access
+        sas_token = generate_blob_sas(
+            account_name=blob_client.account_name,
+            container_name=AZURE_CONTAINER_NAME,
+            blob_name=blob_name,
+            account_key=os.getenv("AZURE_STORAGE_KEY", "ts/PL1cr3X1F9JWgksAtqcWsQvPBK9UJ3BtNQBL98kYU17U3JxEiFI2vJrNDzmAyFRleOdRdoG03+ASt9RDnZA=="),
+            permission=BlobSasPermissions(read=True),
+            expiry=datetime.utcnow() + timedelta(hours=1)
         )
+        blob_url = f"{blob_client_instance.url}?{sas_token}"
 
         background_tasks.add_task(os.remove, local_path)
 
         return {
-            "audio_url": presigned_url,
-            "s3_key": s3_key
+            "audio_url": blob_url,
+            "blob_name": blob_name
         }
     except Exception as e:
         logger.error(f"Error in voice conversion: {e}")

requirements.txt

@@ -20,6 +20,6 @@ funasr==1.1.5
 numpy==1.26.4
 pyyaml
 python-dotenv
-boto3
 uvicorn
 fastapi
+azure-storage-blob